privacy-orchestration
Guides SecretFlow PSI, secure aggregation/comparison, Kuscia, TEEU simulation, and deployment orchestration workflows.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- privacy-orchestration
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Guides SecretFlow PSI, secure aggregation/comparison, Kuscia, TEEU simulation, and deployment orchestration workflows.
- disable-model-invocation
- true
02 资源文件 2
- 参考文档references/privacy-orchestration.md4.6 KB
- 脚本scripts/kuscia_config_smoke.py819 B
03 SKILL.md 正文
Privacy and orchestration
Use this sub-skill when the task depends on privacy-preserving protocols, cluster wiring, or deployment mode selection rather than on direct model or component APIs.
Owns
- PSI helpers and PSI protocol choices
- secure aggregation and comparison helpers
- Kuscia request parsing and SecretFlow cluster conversion helpers
- TEEU simulation guidance and deployment flow
- debug/simulation/production/Kuscia deployment-mode decisions
- common party/device co-location and orchestration failures
Does not own
- plain device setup and data containers — use
runtime-data - direct preprocessing/statistics/classical ML APIs — use
analytics - component CLI / export workflows — use
component-cli
Trigger phrases
Use this route when a user asks things like:
- how to run PSI in SecretFlow
- how to compare or aggregate data securely
- how to parse a Kuscia task config
- how to choose debug, simulation, production, or Kuscia deployment modes
- how to use TEEU in simulation mode
- why a protocol or party layout is failing
Reading order
- Read
references/privacy-orchestration.mdfor the protocol and deployment map. - Read the root troubleshooting page when the problem is an install, backend, or cluster issue that is not unique to this workflow.
- Use
scripts/kuscia_config_smoke.pywhen you want a tiny config-parser check that does not require any external cluster.
Workflow
- Pick the deployment mode before picking a protocol. The runtime shape changes between debug, simulation, production, and Kuscia.
- For PSI or comparison tasks, decide which party owns the result and whether the inputs must already be co-located.
- For Kuscia, parse the request shape first and only then derive the SecretFlow cluster config.
- For TEEU, treat the auth-manager and memory prerequisites as part of the workflow, not as an afterthought.
Common decisions
- Use the simplest PSI protocol that matches the dataset size and privacy needs.
- Use simulation mode when the task is about code flow or basic validation and not about a production deployment.
- Use Kuscia when the task names task orchestration, port reuse, or production scheduling.
- Use TEEU simulation for educational or quick-start validation, not as a production security proof.
Bundled files
references/privacy-orchestration.md— PSI, deployment, Kuscia, and TEEU notes.scripts/kuscia_config_smoke.py— tiny Kuscia config parsing helper.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).